Research On Images’ Characteristics Differences And Fusion Techniques Of Dual-Color Mid-Wave Infrared

Infrared mid-wave sub-band is an important part in multi-color detection. Its images playa very important role both in military and civil fields. By using a dual-color mid-wave imager,two sub-bands images, which are in high quality and have their own advantages, can becollected，but their plenty, accuracy, and reliable target information can be finally got only byfusion. Presently, the researches for dual-color mid-wave image fusion are done usually byprevious assessments. As some fusion algorithms can not be done without some relatedchoices and researches, the improvement of fusion result and efficiency can not be got. Thepaper here is directed to the problem mentioned. It starts with the formation ways of imagedifference characteristics in order to study the fusion ways of the dual-color mid-waveinfrared images.There are mainly three parts in this paper.1. Formation models have been built for the characteristic differences between twosub-bands. The purpose of image fusion is to synthesize the difference information of originalimages. The difference information is an aggregative reflection of various factors which isformed during detecting. Once the imaging characteristic differences change, the image’sdifference characteristics change accordingly, and the same is fusion needs. By deeplyanalyzing the imaging factors in the sub-bands such as radiance, radiance transmission,environment inlfuence, detector response, and so on, the paper tries to research the principlesof imaging difference characteristics. Besides, the imaging evolution models for the infraredpoint target, surface target, and the big target are also built here. The validity of these modelsproposed is proved by the simulation results. The difference rules between the gray scale andthe target background contrast degree for the two sub-band imaging are also brought out,’which produces some instructions for extracting images difference characteristics and making some fusion rules.2.A deep analysis for the difference characteristics between the two sub-bands isshown in this paper. The fusing algorithm should meet the fusion demands of differencecharacteristics among images, then the images, which are both in high quality and ift topeople’s vision, can be formed. So collecting and characterizing difference characteristics isvery important for the research of image fusion. Based on the mapping relations betweeni’maging characteristic differences and image characteristic ones, the images characteristics instatistics, vision, exchange and algebra are shown in the sub-bands. In addition, somemethods such as locally linear embedding(LLE)，radar chart, principal component analysis(PCA)，fractal dimension are used in order to lower the dimension. Then, the image’ssynthetic difference characteristics in the two sub-bands can be shown, which not onlydetermines different fusion technology demands in different imaging conditions，but also laysthe foundation for the study of image fusion.3.Several fusion methods used for dual-color mid-wave infrared images are putforward in this paper，and some fusion images have been got from these methods. Theseimages can show synthetic information from the original images in the two mid-wavesub-bands, and they are more convenient for people to observe them.By setting up the mapping relations between image difference characteristics and fusionstrategies, the foundation for studying the fusion ways is set up. Under the direction of thesemapping relations, two groups of fusion ways are put forward here.(1)Onegroup is for gray image fusion, which includes three ways. The ifrst one is thefusion of segmentation and support value transform, aiming at those images taken under thesun. The second one is the way to combine the dual-tree complex wavelet transform withpossible mapping，aiming at those images taken without the sunlight. The experimental resultsshow that fused images full of information and details. What is more，the operation speed ofthese fusion methods is much faster, which is very helpful to the latter study for fusionautomation and engineering. The last one in this group is the fusion for those dual-colormid-wave infrared weak signal images on the basis of supra-threshold stochastic resonance.The problem of strong noisy images fusion is solved.(2)Theother group of fusion methods is for the color fusion of the dual-color mid-waveinfrared images by using two ways of pseudo-color and color transfer. The pseudo-colorfusion is to combine the local fusion with overall fusion, and to do the gray fusion ifrst, then the color fusion. The pseudo-color fusion can be completed by combining gray fusion basedon support value transform with YCaCp color space mapping. The other way is for colortransfer fusion which is done by using the divergence from both the MWIR1andMWIR2-MWIR1images to increase the U passage and V passage separately, and then, toimprove the fusion result. Some experiments have proved that both subjective evaluation andobjective evaluation for the two fusion methods mentioned above are much better，fusionimages not only have higher resolution, more information and more clearly details, but alsomore closely ift with human perception. The fusion results are helpful to image identificationand understanding, which indicates that the fusion results are significant.